Development of a Neural Network for Target Gas Detection in Interdigitated Electrode Sensor-Based E-Nose Systems

Area of Science:

• * Materials Science and Engineering
• * Electrical Engineering
• * Chemical Sensing Technology

Background:

• * Electronic nose (e-nose) systems are crucial for gas detection.
• * Interdigitated Electrode (IDE) sensors offer a platform for gas sensing.
• * Bioimpedance spectroscopy (BIS) provides a method for measuring sensor responses.

Purpose of the Study:

• * To develop a neural network model for detecting specific gases.
• * To utilize an IDE sensor-based e-nose system with a BIS interface.
• * To evaluate the performance of the system in gas classification.

Main Methods:

• * An e-nose system employing an IDE sensor and BIS interface was constructed.
• * Sensor responses were measured using sinusoidal voltage and processed digitally.
• * Kalman filtering and Principal Component Analysis (PCA) were used for data preprocessing, followed by training a multi-layered neural network with backpropagation and Xavier initialization.

Main Results:

• * The e-nose system demonstrated sensitivity to chloroform (24.86 Ω change) and NO2 (0.7825 Ω/ppm).
• * The neural network achieved an 87.16% test accuracy in classifying NO2, acetone, ethanol, and chloroform.
• * The model was trained rapidly in 239.54 milliseconds, indicating efficient processing.

Conclusions:

• * The developed neural network, integrated with an IDE sensor e-nose, is effective for gas detection.
• * Data preprocessing techniques like Kalman filtering and PCA enhance detection capabilities.
• * Increased sensor sensitivity positively correlates with improved neural network detection performance.